520.490 Analog and Digital VLSI Systems

Fall 2001

Gradient Processor

Justin Chan and Marina Smelyansky
Graduate Advisor: Milutin Stanacevic


Project Overview

Gradient flow is a signal conditioning technique for blind separation and localization of broadband sources impinging on a miniature sensor array [1]. The technique converts the problem of resolving time delayed mixtures to instantaneous mixtures of the time differentiated sources, by evaluating spatial and temporal derivatives on the sensor array.

Implementation of gradient flow requires accurate sensing of small spatial and temporal differentials. This project investigates analog circuit implementation of part of the architecture. The main component of the circuit is a high-gain MOS operational amplifier. 

In our project we have tested for optimal performance of an operational amplifier and laid out a sample circuit.

Gradient Flow Schematics [1]


OpAmp Schematics



Closed-Loop OpAmp Simulation Results




Layout of OpAmp and Gradient Processor


[1] "Gradient Flow Broadband Beamforming and Source Separation," M. Stanacevic, G. Cauwenberghs and G. Zweig, ICA 2001, La Jolla CA, Dec. 2001.